jp-SvetaF: v1.0: jp-SveaF.pt

I've been trying to get this embedding right forever. I'm pleased with the results at this point. This was made from 44 images that I took and does a good job of capturing the likeness. It does not produce outputs of a real person, rather features.

44 images, 12 vectors per token, learning rate of .005, 10,000k steps, to be quite honest I could probably go with a lower learning rate as many produced good likenesses and quality.

I hope you enjoy please review and share your images.